Leading AI-native 6G innovation from concept to patent portfolio

The convergence of AI and telecommunications
Artificial Intelligence (AI) is bringing far-reaching changes to the technologies underpinning the 6th Generation (6G) of cellular networks and our patenting operations, with the patent law being put to unprecedented test (e.g., the patentability of AI-related inventions, and the extent of knowledge of the so-called skilled person). Instead of resisting or disregarding these changes, we prefer to embrace the possibilities that come with AI.
Although we often refer to AI in the context of 6G, the journey of the AI-integration into cellular networks began in 4G, as discussed in my doctoral dissertation, and continues in 5G with AI-enabled solutions e.g., for the air interface. Our initial research results demonstrated substantial network performance gains in terms of throughput (up to 30%), signaling overhead reduction (more than 30%), and energy efficiency (up to 20%) through AI-enabled solutions. And our more recent research found that AI-native 6G will enable devices to adapt to specific conditions experienced in the field, as well as to device and base station-specific hardware impairments. To elaborate more, the AI-native design includes both AI-enabled solutions and AI-enabling mechanisms. The latter span from data collection to efficient protocols for transferring the data consumed and produced by AI. The native integration of AI-enabling mechanisms in 6G will support holistic AI adoption and help us harness the full potential of AI across the network not only to enhance network performance but also to offer AI and data as a service.
AI-related patent protection challenges
We expect that AI-native solutions will form a substantial part of the 6G standards and related Standard Essential Patent (SEP) and implementation patent portfolios. This rapid and disruptive shift is also prompting us to rethink our approach to protecting next-generation wireless innovations, which will be ultimately AI-native.
Common pitfalls in AI-related patent applications include generic formulations, mere use of known AI techniques and insufficient disclosure. In this sense, the European Patent Office (EPO) guidelines specifically require the detailed disclosure of the mathematical methods and the training datasets, when influencing the technical effect, to ensure the technical effect can be reproduced across the claimed scope. For example, barely storing models in a library and selecting one based on measurements would be seen as an obvious design choice rather than an inventive step according to the EPO guidelines. AI-related patent applications must therefore provide the detailed descriptions of training techniques, model architectures, and non-obvious adaptation of AI techniques for specific applications.
To understand it even better, let us look at two recent examples from the EPO case law that highlight the challenges in the AI-related patent protection. In T1191/19, the EPO Board of Appeal refused the patent application due to the insufficient disclosure of training data and neural network structures, and the mere application of a known machine learning technique. Similarly, in T0161/18, the patent application was rejected because it did not disclose suitable input data for training the artificial neural network. These examples further underline the importance of the detailed disclosure in the AI-related patent applications.
Patent portfolio development towards the AI-native 6G
Nokia's portfolio development strategy emphasizes the protection of innovative AI-native 6G solutions while ensuring their compliance with the anticipated 6G standards and fulfilling the AI patentability criteria. This involves identifying the key technology areas for the AI-integration, filing invention disclosures that cover novel AI solutions in these areas, and ensuring that the resulting patent applications provide sufficient technical detail to withstand the scrutiny.
Our portfolio development strategy also highlights the importance of geographical coverage, particularly in key markets where AI-native 6G innovations are expected to shape the next decade. We carefully consider jurisdictional differences in AI patentability criteria and adapt our filing strategies accordingly. This includes tailoring patent applications to address varying requirements for technical character, inventive step assessment, and sufficiency of disclosure across different patent offices.
As also touched upon in my previous blog, creating a high-quality patent portfolio is an elaborate and exhaustive process that requires the work of a ‘well-oiled machine’, where “every cog, gear, and piston operate together in perfect harmony”. Our biggest strength comes from the seamless cooperation and the regular knowledge exchange among our world-class inventors, technical patent specialists, and seasoned patent counsels. We are fearless in testing new ideas and develop new ways of working, and open to continuous learning through collaboration and trainings. All these elements combined will enable us to create the world's best AI-native 6G patent portfolio.